> On Sat, Jun 19, 2010 at 5:11 PM, bsd
> <[email protected]> wrote:
> > So.  They had a developer working on "Polaris" too,
> and axed the project.  Ever hear of the processor
> code-named the "Rock?"  After years and millions, Sun
> also dropped that project.
> >
> > You don't think Oracle would drop an x86 project
> when they clearly want Solaris/SPARC; and couldn't
> care less about Solaris running on x86?
> 
> I'm not sure what you're trying to say -- Rock was
> always a Sparc chip, not x86.
> It was cancelled (according to public sources)
> because after all the
> millions and years spent, it was too hot, too
> power-hungry, and too
> slow for anyone to take seriously (and nothing i've
> heard unofficially
> prior to the buyout contradicts this either).
> 
> The deafening silence from Oracle is annoying (and
> something I think
> they're going to have to change if they want to push
> hardware,
> appliance based or not, but that's a different
> discussion).  What
> things they have said and done to date suggest that
> they're interested
> in things they think will make money, and not
> interested in things
> that won't.  If x86 does it, they'll probably do x86
> (and given the
> fishworks stuff is all x86 based, I'm guessing the
> answer is still
> 'yes'), if sparc does it, I suspect you'll continue
> to see sparc stuff
> as well.

If I were to guess, my guess would be that that's closer to the
mark; they may want it to run on x86 that they sell, and not
care much about that which they don't.  I think that carried to
extremes, that would be a mistake, but not as big a one as dumping
x86 altogether would be, since x86 is the downward price pressure
on just about every other kind of CPU, which makes an x86 port
an insurance policy in case there are one or two more SPARCs
that (like the RK) don't work out at all, or even just don't measure up
(better than SPARC has in some ways for awhile now).  To some
degree, _any_ port also is an insurance policy insofar as it helps
keep platform-specific assumptions well identified and isolated,
so having at _least_ two current ports helps preserve the option of
any future port that unforeseen circumstances may make desirable.

Besides, x86, at least those with sufficient RAS functionality (where
x86 has been catching up some), is fine if not better for some
appliances (esp. if single-thread performance matters).

Some reasons I think that supporting _some_ non-Sun/Oracle hardware
would be  a good idea include: laptops (developers, salespeople,
some others), specialized systems (like really tough hardware for the
military or for industrial control use), and so on. If one can (still?  I think
some were available at times past) buy board sets rather than entire
systems, the latter could still make a bit of money, and both of those
contribute to a vibrant ecosystem.  The biggie for a vibrant ecosystem
though is to be able to run on cheap hardware in colleges and universities,
or even for today's horrible penniless home users, who may well be
making corporate purchasing decisions in a few years (or even now;
I use Solaris at home and at work, in the hopes that much of what
I've learned and continue to learn will be of use in both places).

It ought to be possible to make that sort of thing pay for itself,
where it involves support contracts, and where the seller provides
support for anything that wouldn't arise on a Sun/Oracle platform
(perhaps being responsible for demonstrating that the problem can
also be re-created on such a platform).  Along those lines, it would
be nice to see explicit support for Solaris and OpenSolaris under
specified virtualization products (VirtualBox, maybe xen and VMware)
continue.  And it might be kind of cool to add to VirtualBox a
means of injecting (virtualized) hardware faults, which might help
in both testing and in re-creating some sorts of problems.

That doesn't imply that it makes sense to fund drivers for every last
el cheapo NIC/WiFi/whatever device being sold for x86.  Let the
community (or those that want to claim (Open)Solaris compatibility)
bear the burden of some of those.  How many drivers for hardware
that Sun/Oracle doesn't sell would be worth funding (e.g. for those
laptops) and how many wouldn't is a judgement call, although in
tough times one might suspect that it would be a lot less than
some of us might prefer.  Again, laptops, hardened platforms, and
educational use should almost be viewed as advertising in that
they will to some degree lead to future sales.
-- 
This message posted from opensolaris.org
_______________________________________________
opensolaris-discuss mailing list
[email protected]

Reply via email to